Bits vs Bytes: The Difference Explained

A bit — short for 'binary digit' — is the smallest unit of data: a single 0 or 1. A byte is a group of 8 bits, and those 8 bits can represent 256 different values (2 to the power of 8), which was historically enough for one character of text. Almost everything you store is measured in bytes: a document is so many kilobytes, a photo so many megabytes.

Bits, by contrast, show up mostly when data is in motion. The reason is historical: networks push data along a wire one bit at a time, so engineers naturally counted in bits, while storage groups those bits into addressable bytes.

There is a widely followed shorthand: a capital 'B' means bytes and a lowercase 'b' means bits. So KB is kilobytes while Kb is kilobits; MB/s is megabytes per second while Mbps is megabits per second. The convention is not enforced everywhere, but when you see 'Mbps' with a small b, it is bits — and to get bytes per second you divide by 8.

Keeping this one rule in mind resolves most everyday confusion. If a number looks surprisingly big for a speed or surprisingly small for a file, check the case of the B — it is usually the culprit.

Storage and file sizes are quoted in bytes: your photos, documents, apps and drives. Data transfer and bandwidth are quoted in bits: internet plans, Ethernet (100 Mbps, 1 Gbps), Wi-Fi and mobile speeds. The practical upshot is that whenever you cross from 'how big is it' (bytes) to 'how fast can I move it' (bits), the factor of 8 has to appear.

A worked example ties it together: a 1 gigabyte file is 8 gigabits, so on a 1 Gbps (gigabit) connection it takes about 8 seconds in the best case — not 1 second, as the matching '1 and 1' might tempt you to guess.

For the first two decades of computing a byte was simply "the number of bits needed to hold one character", and that number varied by machine: six bits on some CDC and DEC systems, seven on others, nine on the PDP-10 family. IBM's System/360, announced in 1964, standardised eight bits so that two decimal digits would pack neatly into one byte, and the industry followed because everyone wanted to run IBM's software.

That history is why networking standards say "octet" rather than "byte". When the internet protocols were written the word byte was still ambiguous, so RFCs specify an octet — exactly eight bits, no argument. You will still meet the term in packet documentation and in protocol specifications, and it means precisely what you would expect a byte to mean today.

The division of labour is consistent once you see it. Anything stored is counted in bytes because it is addressed in bytes: file sizes, memory, disk capacity. Anything transmitted is counted in bits because a serial link sends one symbol at a time and has no notion of where a byte begins. Codecs sit on the transmission side of that line, which is why bitrates are quoted in kbps and Mbps.

Line coding complicates the wire side further. Gigabit Ethernet over fibre, SATA and the first two PCIe generations all used 8b/10b encoding, transmitting ten symbols for every eight bits of data to keep the signal balanced and the clock recoverable — a flat 25% overhead, which is why 6 Gbps SATA tops out near 600 MB/s rather than 750. PCIe 3.0 onwards switched to 128b/130b and pays only about 1.5%.